NBER WORKING PAPER SERIES ARTIFICIAL INTELLIGENCE, AUTOMATION AND WORK Daron Acemoglu Pascual Restrepo Working Paper 24196 http://www.nber.org/papers/w24196 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 January 2018 Prepared for Economics of Artificial Intelligence, edited by Ajay Agarwal, Avi Goldfarb and Joshua Gans. We are grateful to David Autor for useful comments. We gratefully acknowledge financial support from Toulouse Network on Information Technology, Google, Microsoft, IBM and the Sloan Foundation. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications. © 2018 by Daron Acemoglu and Pascual Restrepo. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source. Artificial Intelligence, Automation and Work Daron Acemoglu and Pascual Restrepo NBER Working Paper No. 24196 January 2018 JEL No. J23,J24 ABSTRACT We summarize a framework for the study of the implications of automation and AI on the demand for labor, wages, and employment. Our task-based framework emphasizes the displacement effect that automation creates as machines and AI replace labor in tasks that it used to perform. This displacement effect tends to reduce the demand for labor and wages. But it is counteracted by a productivity effect, resulting from the cost savings generated by automation, which increase the demand for labor in non-automated tasks. The productivity effect is complemented by additional capital accumulation and the deepening of automation (improvements of existing machinery), both of which further increase the demand for labor. These countervailing effects are incomplete. Even when they are strong, automation in- creases output per worker more than wages and reduce the share of labor in national income. The more powerful countervailing force against automation is the creation of new labor-intensive tasks, which reinstates labor in new activities and tends to increase the labor share to counterbalance the impact of automation. Our framework also highlights the constraints and imperfections that slow down the adjustment of the economy and the labor market to automation and weaken the resulting productivity gains from this transformation: a mismatch between the skill requirements of new technologies, and the possibility that automation is being introduced at an excessive rate, possibly at the expense of other productivity-enhancing technologies. Daron Acemoglu Department of Economics, E52-446 MIT 77 Massachusetts Avenue Cambridge, MA 02139 and CIFAR and also NBER [email protected] Pascual Restrepo Department of Economics Boston University 270 Bay State Rd Boston, MA 02215 and Cowles Foundation, Yale [email protected] 1 Introduction The last two decades have witnessed major advances in artificial intelligence (AI) and robotics. Future progress is expected to be even more spectacular and many commenta- tors predict that these technologies will transform work around the world (Brynjolfsson and McAfee, 2012; Ford, 2016; Boston Consulting Group, 2015; McKinsey, 2017). Re- cent surveys find high levels of anxiety about automation and other technological trends, underscoring the widespread concerns about their effects (Pew Research Center, 2017). These expectations and concerns notwithstanding, we are far from a satisfactory un- derstanding of how automation in general, and AI and robotics in particular, impact the labor market and productivity. Even worse, much of the debate in both the popular press and academic circles centers around a false dichotomy. On the one side are the alarmist arguments that the oncoming advances in AI and robotics will spell the end of work by humans, while many economists on the other side claim that because technological break- throughs in the past have eventually increased the demand for labor and wages, there is no reason to be concerned that this time will be any different. In this essay, we build on Acemoglu and Restrepo (2016), as well as Zeira (1998) and Acemoglu and Autor (2011) to develop a framework for thinking about automation and its impact on tasks, productivity, and work. At the heart of our framework is the idea that automation and thus AI and robotics replace workers in tasks that they previously performed, and via this channel, create a powerful displacement effect. In contrast to prevailing presumptions in much of macroe- conomics and labor economics, which maintain that productivity-enhancing technologies always increase overall labor demand, the displacement effect can reduce the demand for labor, wages and employment. Moreover, the displacement effect implies that increases in output per worker arising from automation will not result in a proportional expansion of the demand for labor. The displacement effect causes a decoupling of wages and output per worker, and a decline in the share of labor in national income. We then highlight several countervailing forces, which push against the displacement effect and may imply that automation, AI, and robotics could increase labor demand. First, the substitution of cheaper machines for human labor creates a productivity effect: as the cost of producing automated tasks declines, the economy will expand and increase the demand for labor in non-automated tasks. The productivity effect could manifest itself as an increase in the demand for labor in the same sectors undergoing automation or as an increase in the demand for labor in non-automating sectors. Second, capital accumulation triggered by increased automation (which raises the demand for capital) will also raise the demand for labor. Third, automation does not just operate at the extensive margin—replacing tasks previously performed by labor—but at the intensive 1 margin as well, increasing the productivity of machines in tasks that have already been automated. This phenomenon, which we refer to as deepening of automation, tends to create a productivity effect but no displacement, and thus increases labor demand. Though these countervailing effects are important, they are generally insufficient to engender a “balanced growth path,”meaning that even if these effects were powerful, ongoing automation would still reduce the share of labor in national income (and possibly employment which tends to be linked to the labor share). We argue that there is a more powerful countervailing force that increases the demand for labor as well as the share of labor in national income: the creation of new tasks, functions and activities in which labor has a comparative advantage relative to machines. The creation of new tasks generates a reinstatement effect directly counterbalancing the displacement effect. Indeed, throughout history, we have not just witnessed pervasive automation, but a continuous process of new tasks creating new employment opportunities for labor. As tasks in textiles, metals, agriculture and other industries were being automated in the 19th and 20th centuries, a new range of tasks in factory work, engineering, repair, back- office, management and finance generated demand for displaced workers. The creation of new tasks is not an autonomous process advancing at a predetermined rate, but one whose speed and nature are shaped by the decisions of firms, workers and other actors in society, and which might be fueled by new automation technologies. First, this is because automation, by displacing workers, may create a greater pool of labor that could be employed in new tasks. Second, the currently most discussed automation technology, AI itself, can serve as a platform to create new tasks in many service industries. Our framework also highlights that even with these countervailing forces, the adjust- ment of an economy to the rapid rollout of automation technologies could be slow and painful. There are some obvious reasons for this related to the general slow adjustment of the labor market to shocks, for example, because of the costly process of workers being reallocated to new sectors and tasks. Such reallocation will involve both a slow process of searching for the right matches between workers and jobs, and also the need for retraining, at least for some of the workers. A more critical, and in this context more novel, factor is a potential mismatch between technology and skills—between the requirements of new technologies and tasks and the skills of the workforce. We show that such a mismatch slows down the adjustment of labor demand, contributes to inequality, and also reduces the productivity gains from both automation and the introduction of new tasks (because it makes the complementary skills necessary for the operation of new tasks and technologies more scarce). Yet another major factor to be taken into account is the possibility of excessive au- tomation. We highlight that a variety of factors (ranging from a bias in favor of capital in the tax code to labor market imperfections create a wedge between the wage and the 2 opportunity cost of labor) and will push towards socially excessive automation, which not only generates a direct inefficiency but also acts as a drag on productivity growth. Excessive automation could potentially explain why, despite the enthusiastic